The Vocabulary of Agent Governance
A young category runs on borrowed words until someone defines them properly. These are the terms security, compliance, and engineering teams use when AI agents act on real systems: each defined in under 60 words, each backed by a concrete data point, none padded.
AI Agent Governance
The infrastructure that controls what agents may do, enforces it at runtime, and produces verifiable evidence.
Read the definition →Behavioral Firewall
Enforcement between agents and the systems they act on: what is not allowed does not execute.
Read the definition →Observability vs Governance
Telemetry describes behavior after the fact; governance constrains it before execution. The difference, side by side.
Read the definition →AI Agent Audit Trail
A tamper-evident record of every action across six dimensions: who, what, when, where, why, how.
Read the definition →Agent Attestation
Signed, verifiable proof that an agent operated within its mandate, checkable by third parties without raw data.
Read the definition →Trust Score for AI Agents
A behavioral rating of the deployed agent, normalized across providers. Conduct, not capability.
Read the definition →Levels of Agent Autonomy
The ladder from read-only to self-directed, and why promotion should be earned with evidence.
Read the definition →Human-in-the-Loop Approval
Explicit human sign-off before designated actions execute, and why it must be recorded to count.
Read the definition →AI Governance Proxy (AI Gateway)
One network control point for the agent traffic you route through it: identity, policy, evidence, zero code changes.
Read the definition →Knowledge Egress
The continuous outflow of proprietary information through AI usage, invisible to classic DLP.
Read the definition →Definitions are the map. The gateway is the territory.
Start with one process area. Two weeks, a dated readout, and a fixed price in writing.